SuperCLUE-LongContext
Models retrieve information across multiple turns with long Chinese context.
SuperCLUE reports scores in a public workbook.
- Made by
- SuperCLUE
- Kind
- Finding facts in long inputs
- Level
- School
- Scored
- Checked automatically
- Answers
- Short answer
- Measures
- Capability
- Language
- Chinese
- Interaction
- Multi-turn
- Refresh
- One fixed set
- Run by
- A host
- Strongest model tested
- Kimi K3.1 Pro #17 on the capability index
- Last model added
- November 2026
- Scores refreshed
- November 20th, 2026
Qwen4-Max leads with 51.4 Qwen4-Max leads with 51.4
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
- 51.4
Qwen4-Max
- 35.2
DeepSeek V4.1 Flash 1017
- 34.8
Tencent Hy4
- 32.8 Kimi K3.1 Pro
- 31.9
MiniMax-M3.1-Flash
- 26.5 GLM-5.3
- 22.9
Step 5 Preview
- 21.9 MiMo-V2.6 Pro
- 15.3
Qwen3.8 Max (0902)
- 13.1 GLM-5.3-Flash
- 11.0 Doubao-Seed-2.1-Pro-260915
- 10.4
Qwen3.8 Flash
- 9.30 MiMo-V2.6 Flash
- 8.88
LongCat-2.5
- 4.44 Spark-X2.6
- 1
Qwen4-Max 51.4
- 2
DeepSeek V4.1 Flash 1017 35.2
- 3
Tencent Hy4 34.8
- 4 Kimi K3.1 Pro 32.8
- 5
MiniMax-M3.1-Flash 31.9
- 6 GLM-5.3 26.5
- 7
Step 5 Preview 22.9
- 8 MiMo-V2.6 Pro 21.9
- 9
Qwen3.8 Max (0902) 15.3
- 10 GLM-5.3-Flash 13.1
- 11 Doubao-Seed-2.1-Pro-260915 11.0
- 12
Qwen3.8 Flash 10.4
- 13 MiMo-V2.6 Flash 9.30
- 14
LongCat-2.5 8.88
- 15 Spark-X2.6 4.44
Source: Official SuperCLUE-LongContext leaderboard
- 1
Qwen4-Max 51.4
- 2
DeepSeek V4.1 Flash 1017 35.2
- 3
Tencent Hy4 34.8
- 4 Kimi K3.1 Pro 32.8
- 5
MiniMax-M3.1-Flash 31.9
- 6 GLM-5.3 26.5
- 7
Step 5 Preview 22.9
- 8 MiMo-V2.6 Pro 21.9
- 9
Qwen3.8 Max (0902) 15.3
- 10 GLM-5.3-Flash 13.1
- 11
Qwen3.8 Flash 10.4
- 12 MiMo-V2.6 Flash 9.30
- 13
LongCat-2.5 8.88
Source: Official SuperCLUE-LongContext leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 51.4 | Nov 2026 | Official leaderboard | |
| 2 | | 35.2 | Oct 2026 | Official leaderboard | |
| 3 | | 34.8 | Nov 2026 | Official leaderboard | |
| 4 | Kimi K3.1 Pro | 32.8 | Oct 2026 | Official leaderboard | |
| 5 | | 31.9 | Oct 2026 | Official leaderboard | |
| 6 | GLM-5.3 | 26.5 | Aug 2026 | Official leaderboard | |
| 7 | | 22.9 | Sep 2026 | Official leaderboard | |
| 8 | MiMo-V2.6 Pro | 21.9 | Sep 2026 | Official leaderboard | |
| 9 | | 15.3 | Sep 2026 | Official leaderboard | |
| 10 | GLM-5.3-Flash | 13.1 | Aug 2026 | Official leaderboard | |
| 11 | Doubao-Seed-2.1-Pro-260915 | 11.0 | Official leaderboard | ||
| 12 | | 10.4 | Aug 2026 | Official leaderboard | |
| 13 | MiMo-V2.6 Flash | 9.30 | Sep 2026 | Official leaderboard | |
| 14 | | 8.88 | Nov 2026 | Official leaderboard | |
| 15 | Spark-X2.6 | 4.44 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 51.4 | Nov 2026 | Official leaderboard | |
| 2 | | 35.2 | Oct 2026 | Official leaderboard | |
| 3 | | 34.8 | Nov 2026 | Official leaderboard | |
| 4 | Kimi K3.1 Pro | 32.8 | Oct 2026 | Official leaderboard | |
| 5 | | 31.9 | Oct 2026 | Official leaderboard | |
| 6 | GLM-5.3 | 26.5 | Aug 2026 | Official leaderboard | |
| 7 | | 22.9 | Sep 2026 | Official leaderboard | |
| 8 | MiMo-V2.6 Pro | 21.9 | Sep 2026 | Official leaderboard | |
| 9 | | 15.3 | Sep 2026 | Official leaderboard | |
| 10 | GLM-5.3-Flash | 13.1 | Aug 2026 | Official leaderboard | |
| 11 | | 10.4 | Aug 2026 | Official leaderboard | |
| 12 | MiMo-V2.6 Flash | 9.30 | Sep 2026 | Official leaderboard | |
| 13 | | 8.88 | Nov 2026 | Official leaderboard |
Rankings by part
We read 12 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as 64K–128K, 128K–256K and 256K–512K. Qwen4-Max leads 10 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
64K–128K
13 models we track have a score here.
- 1
Qwen4-Max 86.5
- 2
Tencent Hy4 ↑1 75.6
- 3
DeepSeek V4.1 Flash 1017 ↓1 62.9
- 4 GLM-5.3 ↑2 61.9
- 5
MiniMax-M3.1-Flash 57.1
- 6 MiMo-V2.6 Pro ↑2 40.5
- 7
Qwen3.8 Max (0902) ↑2 31.1
- 8 Kimi K3.1 Pro ↓4 27.0
- 9
Step 5 Preview ↓2 25.6
- 10
Qwen3.8 Flash ↑1 21.5
128K–256K
13 models we track have a score here.
- 1
Qwen4-Max 75.6
- 2
DeepSeek V4.1 Flash 1017 49.9
- 3
MiniMax-M3.1-Flash ↑2 42.5
- 4 Kimi K3.1 Pro 41.9
- 5
Tencent Hy4 ↓2 33.1
- 6
Step 5 Preview ↑1 31.5
- 7 GLM-5.3-Flash ↑3 29.4
- 8 GLM-5.3 ↓2 17.9
- 9
LongCat-2.5 ↑4 15.5
- 10
Qwen3.8 Max (0902) ↓1 15.1
256K–512K
13 models we track have a score here.
- 1
Qwen4-Max 45.2
- 2 Kimi K3.1 Pro ↑2 41.1
- 3
MiniMax-M3.1-Flash ↑2 32.9
- 4
DeepSeek V4.1 Flash 1017 ↓2 30.4
- 5
Tencent Hy4 ↓2 27.5
- 6 MiMo-V2.6 Pro ↑2 23.6
- 7
Step 5 Preview 21.6
- 8 GLM-5.3 ↓2 18.2
- 9
Qwen3.8 Max (0902) 17.8
- 10 GLM-5.3-Flash 11.2
512K–1M
13 models we track have a score here.
- 1 Kimi K3.1 Pro ↑3 24.1
- 2
Qwen4-Max ↓1 20.7
- 3
Tencent Hy4 17.1
- 4 GLM-5.3 ↑2 17.1
- 5
Step 5 Preview ↑2 17.1
- 6 MiMo-V2.6 Pro ↑2 14.7
- 7
DeepSeek V4.1 Flash 1017 ↓5 13.7
- 8
MiniMax-M3.1-Flash ↓3 9.65
- 9 MiMo-V2.6 Flash ↑3 7.14
- 10
LongCat-2.5 ↑3 4.60
原文复刻
13 models we track have a score here.
- 1
Qwen4-Max 52.8
- 2
Tencent Hy4 ↑1 42.4
- 3 MiMo-V2.6 Pro ↑5 29.9
- 4 GLM-5.3 ↑2 27.9
- 5
Step 5 Preview ↑2 17.9
- 6
DeepSeek V4.1 Flash 1017 ↓4 17.9
- 7
MiniMax-M3.1-Flash ↓2 17.8
- 8
Qwen3.8 Flash ↑3 9.28
- 9 Kimi K3.1 Pro ↓5 8.04
- 10 GLM-5.3-Flash 6.89
句后插入
13 models we track have a score here.
- 1
Qwen4-Max 63.1
- 2 Kimi K3.1 Pro ↑2 53.4
- 3
MiniMax-M3.1-Flash ↑2 53.1
- 4 MiMo-V2.6 Pro ↑4 52.9
- 5
Tencent Hy4 ↓2 45.9
- 6
DeepSeek V4.1 Flash 1017 ↓4 42.4
- 7
Step 5 Preview 40.2
- 8 GLM-5.3 ↓2 24.1
- 9
Qwen3.8 Flash ↑2 17.9
- 10
Qwen3.8 Max (0902) ↓1 14.6
多段提取拼接
13 models we track have a score here.
- 1
Qwen4-Max 73.0
- 2
DeepSeek V4.1 Flash 1017 53.5
- 3
Tencent Hy4 52.2
- 4 Kimi K3.1 Pro 40.8
- 5 GLM-5.3 ↑1 32.7
- 6
MiniMax-M3.1-Flash ↓1 29.4
- 7 MiMo-V2.6 Pro ↑1 28.9
- 8 GLM-5.3-Flash ↑2 20.3
- 9
Qwen3.8 Max (0902) 17.8
- 10
Qwen3.8 Flash ↑1 14.4
条件筛选
13 models we track have a score here.
- 1
Qwen4-Max 43.0
- 2 Kimi K3.1 Pro ↑2 38.5
- 3
DeepSeek V4.1 Flash 1017 ↓1 33.4
- 4
Tencent Hy4 ↓1 27.1
- 5
MiniMax-M3.1-Flash 27.0
- 6 GLM-5.3 24.7
- 7
Step 5 Preview 23.2
- 8
Qwen3.8 Max (0902) ↑1 16.1
- 9 MiMo-V2.6 Pro ↓1 15.9
- 10 GLM-5.3-Flash 14.6
关系定位
13 models we track have a score here.
- 1
Qwen4-Max 47.4
- 2
MiniMax-M3.1-Flash ↑3 35.5
- 3
DeepSeek V4.1 Flash 1017 ↓1 30.9
- 4
Tencent Hy4 ↓1 30.9
- 5 GLM-5.3 ↑1 26.1
- 6
Step 5 Preview ↑1 24.1
- 7 Kimi K3.1 Pro ↓3 23.4
- 8 MiMo-V2.6 Pro 15.9
- 9
Qwen3.8 Max (0902) 15.5
- 10 MiMo-V2.6 Flash ↑2 12.8
8 needles
13 models we track have a score here.
- 1
Tencent Hy4 ↑2 73.0
- 2
Qwen4-Max ↓1 59.9
- 3
Step 5 Preview ↑4 52.8
- 4 MiMo-V2.6 Pro ↑4 49.2
- 5 Kimi K3.1 Pro ↓1 49.0
- 6
DeepSeek V4.1 Flash 1017 ↓4 45.0
- 7
MiniMax-M3.1-Flash ↓2 37.9
- 8 GLM-5.3 ↓2 35.0
- 9 GLM-5.3-Flash ↑1 25.5
- 10
Qwen3.8 Flash ↑1 23.1
12 needles
13 models we track have a score here.
- 1
Qwen4-Max 53.4
- 2 GLM-5.3 ↑4 34.8
- 3
Tencent Hy4 32.8
- 4
MiniMax-M3.1-Flash ↑1 31.3
- 5
DeepSeek V4.1 Flash 1017 ↓3 31.1
- 6 Kimi K3.1 Pro ↓2 26.8
- 7
Step 5 Preview 19.0
- 8
Qwen3.8 Max (0902) ↑1 16.6
- 9
LongCat-2.5 ↑4 13.3
- 10 MiMo-V2.6 Pro ↓2 12.0
16 needles
13 models we track have a score here.
- 1
Qwen4-Max 48.4
- 2
DeepSeek V4.1 Flash 1017 35.3
- 3 Kimi K3.1 Pro ↑1 32.4
- 4
MiniMax-M3.1-Flash ↑1 30.9
- 5
Tencent Hy4 ↓2 27.4
- 6 MiMo-V2.6 Pro ↑2 21.3
- 7 GLM-5.3 ↓1 20.2
- 8
Step 5 Preview ↓1 18.4
- 9
Qwen3.8 Max (0902) 14.2
- 10 GLM-5.3-Flash 11.3
3 models have held number one on the official leaderboard since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official SuperCLUE-LongContext leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official SuperCLUE-LongContext leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Long context
- MRCR v2 In use Multi-round coreference: retrieve the n-th of several similar requests hidden in a long synthetic conversation, up to 1M tokens.
Muse Spark 1.3 98.5%
- LongBench v2 In use 503 multiple-choice questions over 8K-2M-word contexts (QA, dialogue, code repos, structured data). Accuracy.
Inkling 2 Small 70.6%
- GraphWalks In use A graph is given as an edge list in long context. The model answers BFS or parent queries, scored by F1. Claude Fable 5 99.96%
- OpenAI MRCR v2 In use Models read a long synthetic chat that hides 2, 4 or 8 identical writing requests and must reproduce the response to the i-th one. Scored by mean string match ratio per context bin.
Gemini 3.7 Flash 89%
- CL-bench In use Model must learn new knowledge from a supplied long context (rules, procedures, data) and solve tasks. Strict LLM-verified rubrics. GPT-5.4 27.9%
- AA-LCR 1.1 In use Models answer 100 questions that need reasoning across sets of real documents of about 100K tokens. An LLM equality checker with a fixed system prompt compares each answer with the key.
Step 5 Preview 88.3%
Data audit
8 of 9 checks pass, 1 warning. Look first at the Task access check. We don’t know whether the tasks are public. Every benchmark
Catalog record 1 warning 1 pass
- Warning Task access We don’t know whether the tasks are public.
- Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
- Pass Independent runs Scores from the official leaderboard (15 entries).
- Pass Models scored 13 models we track, plus 2 other entries on its leaderboard.
- Pass Newest score The newest is from November 2026, for Qwen4-Max.
- Note Frontier models None of the top 10 has a score. The strongest tested is Kimi K3.1 Pro (#17).
- Note Headroom Scored as a plain number, with no fixed ceiling.
Leaderboard import 3 pass
- Pass Importer We read the board on superclueai.com: 13 rows for models we track and 2 other entries at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.